JP2003280922A - Code size decreasing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve such a problem in a prior art that the program size becomes large due to a large size of a code because a prior compiler outputs a function body whenever a symbol of a function is referenced, even for a function which is never called directly nor indirectly from a program as a result of optimization such as in-line development. <P>SOLUTION: The function body which is never called directly or indirectly from a program and which has a symbol referred to from a program is deleted after the function symbol is rewritten into a specific constant number for the function. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a programming language processing technology.

[0002]

2. Description of the Related Art One method of reducing the code size is to delete the code of a function that is never called.

[0003]

In realizing an object-oriented programming language, function symbols may be referred to in addition to function calls. Object-oriented programming provides a function called virtual function call. Its simple realization is a function call via a dispatch table. In order to speed up the function call, I-call if conversion (method test conversion, etc.) may be performed. Reference: "Detref
s, et. al. : Inlining of Virtu
al Methods, "ECOOP99, pp.25.
8-278).

To further speed up the function call,
In-line deployment (Reference: AV Aho & R Cecy,
"Compiler technique and principle", Science Co.). This is a technology that directly expands the function body to the calling part.

A specific example of a method for implementing a virtual function call will be shown. FIG. 1 shows an example of indirect call. The address of the target function is acquired from the dispatch table (see FIG. 2), which is the data structure representing the address of each function of the source code generated by the compiler, and the function is called.
FIG. 3 shows an example of method test conversion. If the address of the target function is acquired from the dispatch table and the address is the function methodA, the function met
Call hodA directly. It can be seen from FIGS. 2 and 3 that the function symbol is referenced from the code of the virtual function call optimized by the dispatch and the method test conversion.

Since the inline expansion which is a conventional technique has only the effect of deleting the function symbol appearing at the caller of the function call, the function symbol appearing in the dispatch table or the code of the method test conversion in terms of implementing these virtual function calls. Cannot be deleted. At this time, the function code in which the function symbol remains cannot be deleted even if there is no possibility that the function code will be called even if all the functions are inline expanded to the caller.

[0007]

[Means for Solving the Problems] A function which is not called directly or indirectly but whose function name is referred to as an address is obtained, and a reference to the function symbol is replaced with a constant unique to the function. . Similarly, instead of registering the address of the function (that is, describing the function name) in the method dispatch table, a unique constant representing the target function is registered. By doing so, since the reference to the function name (symbol) disappears, the function body after inline expansion can be deleted and the code size can be reduced.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings. The present invention is not limited to the examples.

FIG. 4 shows a computer on which the compiler of the present invention operates. The computer has a processor 401 and a main memory 402,
It has a bus 403, a disk device 404, and a CD-ROM reading device 405. The compiler is stored in the CD-ROM (storage medium) 406. Computer is CD-
When the CD-ROM 406 is set in the ROM reading device 405, the compiler program is read and stored in the disk device 404 via the bus 403. When executing the compiler, the compiler program is loaded from the disk device 404 to the main memory 402 via the bus 403.
, And the processor 401 reads out and executes at any time.
Although not shown, the computer is connected to an input device for inputting a compiler execution instruction and the like, a display device for displaying the compiler execution result, and a printing device for printing the compiler execution result.

An embodiment of the compiler of the present invention is shown in FIG. The compiler 501 receives the source program 502 as an input, parses the intermediate code conversion unit 503 to convert it into intermediate code, and the optimization unit 504 I-cal.
Optimization such as l if conversion and inline expansion is performed, the code of the deletable function is deleted by the redundant code deletion unit 505, and finally the intermediate code is converted into the machine code 507 by the code generation unit 506. In the present embodiment, the redundant code deletion unit 505 also detects a function that is not a direct call target and deletes the code of that function.

As shown in FIG. 6, the redundant code deleting unit 505 of FIG. 5 sets a mark unit 601 for setting a flag indicating whether or not each function can be deleted, and deletes the function if it can be determined by checking the flag. It is divided into sweep sections.

Details of the mark portion 601 of FIG. 6 are shown in FIG. First, the mark unit 601 stores all the sentences of all programs configured by the application in the set S. Then, a flag indicating the calling status is set to all functions. 4 for the flag
Three types, UNKNOWN, INDIRECT, DIR
There are ECT and ADDRESS. Each flag is
UNK indicating that no code is called
It means NOW, INDIRECT indicating indirect call, DIRECT indicating direct call, and ADDRESS indicating that only the symbol of the function is referenced.

First, the flags of all functions are set to UNKNOW.
Set to N. If the set S is the empty set φ in the decision 702, the process is terminated. If it is not the empty set φ, one sentence from the set S
Take out one. Then, the process proceeds to decision 704, and if the retrieved statement is an indirect call, the process proceeds to processing 705. In process 705, IND the flags of all functions that can be indirectly called by the statement.
Set to IREC. In decision 704, if the statement is not an indirect call, the procedure proceeds to decision 706, it is determined whether the statement is a direct call, and if it is a direct call, the procedure proceeds to decision 707. In the judgment 707, it is judged whether the flag of the callee function directly called by the statement is ADDRESS or UNKNOWN, and if it is ADDRESS or UNKNOWN, the flag of the function is set to DIRECT. In decision 707, if the statement is not a direct call, the procedure proceeds to decision 709, it is determined whether the statement is a reference to a function symbol, and if it is a reference to a function symbol, the procedure proceeds to decision 710. Judgment 71
In 0, the flag of the function that the statement refers to is UNKNO
It is determined whether it is WN, and if it is UNKNOWN, the function is set to ADDRESS. When the judgment 709 is removed, the process returns to the judgment 702.

If the set S is an empty set φ in the decision 702, FIG.
The mark portion 601 of 1 is finished, and the process proceeds to the sweep portion 602.
In the sweep unit 602, the function flag is ADDRESS.
Alternatively, the function that is UNKNOWN is targeted for deletion, and the deletion is performed. If the function flag is UNKNOWN, it means that the function symbol was not referenced at all.
Delete the function. If the flag of the function is ADDRESS, there is a symbol reference to the function, but since there is no caller that directly or indirectly calls the function, the function is deleted similarly to the flag UNKNOWN. In the compiler of the present invention, the address of the function, that is, the start address of the machine code that implements the function is aligned on a 4-byte boundary, and therefore the function address is an even number. Therefore, in order to avoid the unique constant from matching other function addresses, an odd number is used as the unique constant. Therefore, the optimization of FIG. 2 results in FIG. FIG. 8 shows that when the flag of the 0th function on the function dispatch table is ADDRESS, the function address is replaced with a unique constant “1”. When a plurality of ADDRESS is set as the function flag, the sweep unit assigns a constant other than the unique constant that was previously replaced (in this case, “1” for 0x00000000) and assigns the same constant to different functions. Works like no. It is desirable to manage the relationship between the function replaced by the sweep section and the unique constant so that the correspondence table can be output if necessary.

[0015]

As a result of optimization such as inline expansion, the function is no longer called directly or indirectly, but for functions in which only the function symbol is referenced as an address, the reference to the function symbol is unique to the function. It is possible to delete the function body by replacing it with another constant. As a result, the code size can be reduced. According to the present invention, the code size can be generally reduced by about several percent.

[Brief description of drawings]

FIG. 1 is a diagram showing an indirect call example.

FIG. 2 is a diagram showing a data structure representing an address of a function.

FIG. 3 is a diagram showing a method test conversion example.

FIG. 4 is a diagram showing a configuration of a computer that executes a compiler of the present invention.

FIG. 5 is a flowchart showing the flow of the compiler of the present invention.

FIG. 6 is a detailed flowchart of a deletion method search.

FIG. 7 is a detailed flowchart of a deletion method search.

FIG. 8 is a diagram showing an example of a data structure representing an address of a function according to the present invention.

[Explanation of symbols]

401: processor, 402: main storage device, 403: bus, 404: disk device, 405: CD-ROM reading device, 406: CD-ROM.

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Ryutaro Takahashi             6-81 Onoe-cho, Naka-ku, Yokohama-shi, Kanagawa             Hitachi Software Engineering Stock Association             In-house (72) Inventor Yuji Chiba             1099 Ozenji, Aso-ku, Kawasaki City, Kanagawa Prefecture             Ceremony company Hitachi Systems Development Laboratory F-term (reference) 5B081 AA09 CC16 CC24

Claims (4)

[Claims] 1. In the process of compiling a program written in a programming language that provides virtual function calls,
A code size reduction method for detecting a deletable function from functions that are not called and deleting the code of the function. 2. A statement in a program is scanned and a function whose function name is referred to as an address is not directly or indirectly called, but a reference to the function symbol is made unique to the function. The code size reduction method according to claim 1, wherein the code of the function is deleted after replacing with a constant. 3. A step of scanning a statement in a program, a step of obtaining a function that is not directly or indirectly called but a function name is referred to as an address, and a reference to the function symbol is performed. Replacing with a constant specific to the function, and then deleting the code of the function. 4. A step of scanning a statement in a program, a step of obtaining a function that is not directly or indirectly called but a function name is referred to as an address, and a reference to the function symbol is performed. A storage medium readable by an information processing apparatus, which stores a compiler having a step of deleting a code of the function after substituting with a constant unique to the function.